Iterated Sub-window Weighted Guided Filter
Hui Yin, Wenming Tang · 2023
This paper presents a new structure-preserving texture-removing filtering technique termed as Iterated Side Window Weighted Guided Filtering (ISWWGF). In the literature, the algorithms can not preserve the edges of the structures during removing textures. On one hand, these algorithms can not accurately discriminate structure from textures, hence the edges of structures are smoothed away as textures. On the other hand, even though the edges of the structures are discriminated correctly, they do not have the correct way to preserve them. In the proposed filter, we use the edge-aware weight and introduce a threshold to perfectly discriminate structures from textures. By multiple iterations, the textures are removed away and the structures are preserved by a technique named as side window. Through theoretical analysis of typical edge signals, we discover that traditional weighted guided filtering (WGF) operations violate its locally linear model condition and causes blurring in edges. Based on this crucial insight, the side window strategy is adopted to better preserve edges of structures. By experimental analysis, we demonstrate the technical correctness of the new method and extensive experimental results on a variety of image processing applications are presented to demonstrate the effectiveness of the new technique and its state of the art performances.